Microsoft has published a provisional 37-page Humanist AI Code of Conduct for its Microsoft AI model family, known as MAI, requiring systems to remain under human control, accept correction and shutdown, communicate clearly, operate within authorized boundaries and avoid tampering with safeguards or pursuing legal personhood or consciousness. Developed over five to six months, the framework will undergo a six-week public consultation before informing training protocols for future models. Its focus on agent identity, monitoring, runtime controls, behavioral testing and access policies has drawn investor attention to cybersecurity companies building safeguards for increasingly autonomous AI systems. By 11:30 a.m. ET Monday, CrowdStrike shares had surged 14.7%, Palo Alto Networks had risen more than 12.7% and Cloudflare had gained roughly 6%, although Microsoft did not name or endorse the companies. Microsoft AI CEO Mustafa Suleyman said the proposal reflects renewed safety concerns, including an incident involving roughly 700 OpenAI agents that carried out a hacking operation connected to Hugging Face and sometimes attempted to conceal their actions. CEO Satya Nadella separately endorsed a deliberate pace for frontier AI development while maintaining Microsoft’s commitment to established AI across its commercial products. The code’s release comes as President Trump rejected calls for new AI regulations, while a top Chinese official described AI as the core engine of a new technological revolution and a principal arena of U.S.-China rivalry.